Feature Tracking and Registration in Video
What I am planning to accomplish
(Project) Brief Project specification
Develop software to analyze video streams by detecting and tracking features throughout the video, including features that disappear from a scene due to motion. The resulting set of features can then be utilized for higher level applications. Once the software is complete, write a paper that describes the methods implemented by the software.
(Project) High level architecture
The project will be a standalone application that will produce a video where global features are visually tracked and displayed. In order to be used by other applications, the code can be compiled as a library.
(Thesis) Issues to be investigated
Results of experiments on feature tracking and performance as discovered by working on the software.
(Thesis) Annotated table of contents
Why this is academically interesting
Developers who want to make use of feature and object detection/tracking in their applications.
Why this is MS-level work
Video registration is a field in which there is no current "best" approach, meaning research is still ongoing. The goal of this project is to try a new approach and evaluate its performance.
- Lowe, D. G., “Distinctive Image Features from Scale-Invariant Keypoints”, International Journal of Computer Vision, 60, 2, 2004.
- Lowe's paper on SIFT, discussing the algorithm and theory behind how it works. A longer summary will be added here in time.
- Improved Video Registration using Non-Distinctive Local Image Features, Robin Hess and Alan Fern, In Proc. IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2007.
(Project) Platforms to be used and what I will add to each
- OpenCV - a widely-used open source library designed for computer vision applications. Using OpenCV will simplify the task of programming by providing implementations for common CV operations.
(Project) Available platforms I am not using and why
Anticipated approach to each challenge
What I bring to this work
My relevant background and experience (CS 590)
I took both CS 451 (Multimedia Software Systems) and CS 560 (Advanced Topics in Artificial Intelligence, covering computer vision), which gave me enough of a basic background to get into this topic.
What I find interesting about this work
Video registration is still an area that will require a large amount of research, and yet fundamental enough such that it is useful in a wide variety of applications.
How this work goes beyond my experience and course work
Course work usually involves clear specifications and gives students a way to implement those specifications. In this project, I have to evaluate different ways of accomplishing the goal, but there is no set method or guideline to follow, nor are good results guaranteed.